Acoustic Emission and Machine Learning for Smart Monitoring of Cable Damages in Bridges

This article discusses the use of acoustic emission and machine learning tools for smart monitoring of cable damages in bridges. The need for discovering and measuring the degradation of metallic bridges cables comes out as a must for users’ safety, transportation facility and economic evidence. And the application of acoustic emission for this purpose is emphasized since 1969. However, using this approach to diagnose civil engineering cables is scarce and previous research on friction in the industry and on damaged cables has shown that acoustic emission (AE) signals depend on materials, surface condition, pressure, and relative velocity. Therefore, separating the different sources of acoustic emissions recorded during cable monitoring to identify the signals caused by wire breakages remains a significant challenge. The study focuses on investigating intrafilamentary friction by leveraging the differences in acoustic signatures between cables with broken wires and intact cables. To achieve that, the article suggests an original experimental approach combined to machine learning algorithms to isolate AE sources recorded during data collection. After the experimental setup, parametric analysis, clustering, and classification techniques have been employed to separate different AE sources. The proposed approach could enable tracking the evolution of friction from broken wires over time and recognizing wire break signals even without a reference state for the cable.